liaozhaorun
a9e4746239
- 修复 data_loader.py 财务数据日期过滤，支持按范围加载 - 优化 MADClipper 使用窗口函数替代 join，提升性能 - 修复训练日期边界问题，添加1天间隔避免数据泄露 - 新增 .gitignore 规则忽略训练输出目录
31 lines
789 B
Python
31 lines
789 B
Python
"""训练流程入口脚本
|
||
|
||
运行方式：
|
||
uv run python -m src.training.main
|
||
|
||
或：
|
||
uv run python src/training/main.py
|
||
"""
|
||
|
||
from src.training.pipeline import run_training
|
||
|
||
|
||
if __name__ == "__main__":
|
||
# 运行完整训练流程
|
||
# 训练集：20190101 - 20231231
|
||
# 验证集：20240102 - 20240531 (与训练集间隔1天，避免数据泄露)
|
||
# 测试集：20240602 - 20241231 (与验证集间隔1天，避免数据泄露)
|
||
result = run_training(
|
||
train_start="20190101",
|
||
train_end="20231231",
|
||
val_start="20240102",
|
||
val_end="20240531",
|
||
test_start="20240602",
|
||
test_end="20241231",
|
||
top_n=5,
|
||
output_path="output/top_stocks.tsv",
|
||
)
|
||
|
||
print("\n[Result] Top stocks selection:")
|
||
print(result)
|